JavaScript Engine in Rust

A JavaScript engine written in Rust, with a full pipeline from source text to bytecode execution:

Source -> Lexer -> Parser -> Compiler -> VM

The repository contains:

• A library crate: javascript
• A CLI binary crate: js
• Rust integration tests and JavaScript self-tests
• A Node-based Test262 runner under ci/runner.js

Current status

This project is actively developed and already passes a large portion of Test262 in practice.

Important: two well-known missing areas are still:

• decorators
• proper tail calls (tail call optimization)

If you need either today, use transpilation (for decorators) or avoid relying on spec-level PTC behavior.

Language mode policy

This engine is intentionally not a legacy-web-quirk-compatible sloppy-mode runtime.

• This engine does not explicitly target full, historical sloppy-mode compatibility.
• Legacy web quirks and edge-case loose-mode behaviors are not a primary compatibility goal.
• In practice, strict and modern ECMAScript behavior is prioritized over preserving legacy oddities.

Highlights

• Bytecode VM architecture (not AST tree-walking execution)
• ECMAScript parser and compiler implemented in Rust
• Supports script and module execution
• Includes async-related features (Promises, async/await, top-level await in module mode)
• REPL with persistent command history
• Cross-platform CLI (Linux, macOS, Windows)

Architecture

Core source layout:

• src/core/tokenizer.rs: lexer/tokenizer
• src/core/parser.rs: parser (AST construction)
• src/core/statement.rs: AST node definitions
• src/core/compiler.rs: AST to bytecode compilation
• src/core/opcode.rs: opcode definitions
• src/core/vm.rs and src/core/vm/: VM runtime and builtins
• src/core/value.rs: runtime value model
• src/core/mod.rs: top-level compile/run pipeline and public core entry points
• src/lib.rs: crate public API re-exports
• js/src/main.rs: CLI entry point and REPL loop

Build and run

Build

cargo build --all-features --release
cargo build -p js --release

Run CLI

# Run a file
cargo run -r -p js -- path/to/script.js

# Run a module
cargo run -r -p js -- --module path/to/module.js
cargo run -r -p js -- path/to/module.mjs

# Evaluate inline code
cargo run -r -p js -- -e "1 + 2"

# Start REPL
cargo run -r -p js

CLI options

Usage: js [OPTIONS] [FILE]

Options:
  -e, --eval <EVAL>
      --timer-wait-ms <TIMER_WAIT_MS>
      --module
  -h, --help
  -V, --version

REPL behavior

• Exit with .exit or Ctrl-D
• Multiline input is supported for incomplete code
• History is stored at ~/.js_repl_history (resolved via dirs::home_dir())

Library usage

Add dependency:

[dependencies]
javascript = "0.1.14"

Basic script evaluation:

use javascript::evaluate_script;

fn main() {
    let out = evaluate_script("1 + 2", false, Option::<&std::path::Path>::None).unwrap();
    assert_eq!(out, "3");
}

Module evaluation:

use javascript::evaluate_script;

fn main() {
    let src = "export const x = 1;\n x";
    let out = evaluate_script(src, true, Option::<&std::path::Path>::None).unwrap();
    println!("{out}");
}

Persistent REPL API:

use javascript::Repl;

fn main() {
    let mut repl = Repl::new();
    repl.eval("let a = 10;").unwrap();
    let out = repl.eval("a + 5").unwrap();
    assert_eq!(out, "15");
}

Other exported APIs include tokenize, parse_statement, parse_statements, read_script_file, and value/string helpers.

Testing and quality checks

Rust tests

cargo test -r --all-features --tests

Lint and format

cargo clippy -r --all-features --all-targets -- -D warnings
cargo fmt --all

JS self-test sweep

# Uses bash + timeout logic and categorizes PASS/FAIL/SKIP/TIMEOUT
bash ./test-vm.sh

Windows note: run through bash (for example Git Bash), not directly as a PowerShell script.

Test262 runner

Node-based runner:

# Requires Node.js
node ci/runner.js --limit 10000 --focus "language/expressions/addition"

# Run one focused case
node ci/runner.js --limit 1 --focus "language/literals/string/S7.8.4_A4.1_T1.js"

Key runner options:

• --focus (required, supports multiple and comma-separated values)
• --limit
• --jobs
• --timeout
• --fail-on-failure
• --keep-tmp

The runner builds the engine, composes harness files, runs tests, and writes logs to test262-results.log.

Known gaps

As of current implementation:

• decorators: not implemented
• tail-call-optimization: not implemented
• full historical sloppy-mode quirks: intentionally not a compatibility target

These are common gaps across many real-world engines/runtimes as well, but they remain spec-visible differences.

Contributing

1. Run format, clippy, and tests before opening PRs.
2. Keep parser/compiler/vm changes accompanied by tests (Rust and/or JS scripts).
3. For conformance work, include focused Test262 repro paths in commit or PR notes.

License

MIT. See LICENSE.